9 demos of Gemini Omni and Gemini 3.5 in action
Google shows how Gemini Omni can edit and generate video, while Gemini 3.5 Flash is aimed at agentic workflows and coding.
Intelligence analysis by GPT-5.4 Mini
Google uses nine demos to frame Gemini Omni as a video-first creative model and Gemini 3.5 Flash as a faster engine for long-horizon agent tasks, coding, and custom Search experiences.
Google is showing off two new AI helpers. One is good at making and changing videos by following simple spoken instructions. The other is good at doing long jobs step by step, like a careful assistant that keeps working until the task is finished.
One part is like a video painter. It can take a clip and change what is happening in it, like turning a statue into bubbles or moving a violinist to a new place.
The other part is like a tiny robot office helper. It can sort files, help build web pages, and make Search answer questions in more useful ways. Google says these tools will show up in its apps and services.
Analysis
At Google I/O 2026, Google introduced Gemini Omni and the Gemini 3.5 family, then used this post to show what the models are meant to do. Omni is presented as a multimodal model that can take images, audio, video, and text and use that input to generate videos grounded in Gemini’s knowledge. Google also says it can edit video through conversation, with each instruction building on the last while keeping characters, scene continuity, and physical behavior consistent.
Gemini Omni
The demos focus on editing as much as generation. Google shows prompts that change a sculpture into bubbles, shift a scene’s action, or repeatedly refine a clip across multiple turns. One example starts with a violinist and then moves the musician into a different environment, makes the violin disappear, and changes the camera angle. The framing is that the video becomes raw material for a new scene rather than a fixed recording.
Gemini 3.5 Flash
Google describes 3.5 Flash as the first model in the 3.5 line and emphasizes speed plus performance for agentic work. The post says it can handle long-horizon tasks, multi-step workflows, and coding, especially when paired with the updated Antigravity harness. Google also uses it to show examples like automatically renaming and categorizing messy assets, generating different UI approaches in AI Studio, and powering Search experiences that can produce custom generative UI, dashboards, trackers, mini apps, and information agents.
The rollout details are broad: Omni is going out through the Gemini app, Google Flow, YouTube Shorts, and YouTube Create, with APIs to follow; 3.5 Flash is available through Antigravity, AI Studio, Android Studio, Gemini Enterprise products, AI Mode in Search, and the Gemini app. The post is less a research paper than a product map for where Google wants these models to show up next.
Key points
- Google announced Gemini Omni and Gemini 3.5 at I/O 2026 and used nine demos to show them off.
- Omni is designed to generate and edit video from multimodal inputs such as images, audio, video, and text.
- Gemini 3.5 Flash is positioned for agentic workflows, coding, and other long-horizon tasks.
- Google says 3.5 Flash powers new Search features, custom UI, and information agents.
- The post lists broad rollouts across the Gemini app, Flow, Search, AI Studio, Android Studio, and enterprise tools.
